autolabel,nuscenes: fixed autolabeling token

autolabel.py

@@ -1,5 +1,4 @@
 import argparse
-import importlib
 import os
 from multiprocessing.pool import ThreadPool
 from typing import Dict
@@ -19,7 +18,7 @@
 from torch.utils.data.distributed import DistributedSampler
 
 from torchdrive.data import Batch, TransferCollator
-from torchdrive.datasets.autolabeler import LabelType, save_tensors
+from torchdrive.datasets.autolabeler import AutoLabeler, LabelType, save_tensors
 from torchdrive.datasets.dataset import Dataset
 from torchdrive.train_config import create_parser, TrainConfig
 
@@ -31,8 +30,7 @@
 
 
 # pyre-fixme[5]: Global expression must be annotated.
-config_module = importlib.import_module("configs." + args.config)
-config: TrainConfig = config_module.CONFIG
+config: TrainConfig = args.config
 
 # overrides
 config.num_frames = 1
@@ -56,6 +54,9 @@
 
 dataset: Dataset = config.create_dataset(smoke=args.smoke)
 
+if isinstance(dataset, AutoLabeler):
+    dataset = dataset.dataset
+
 sampler: DistributedSampler[Dataset] = DistributedSampler(
     dataset,
     num_replicas=WORLD_SIZE,

torchdrive/datasets/nuscenes_dataset.py

@@ -361,7 +361,7 @@ def _getitem(self, sample_data: SampleData) -> Dict[str, object]:
             "color": img,
             "mask": None,
             # pyre-fixme[27]: TypedDict `SampleData` has no key `sample_token`.
-            "token": sample_data["sample_token"],
+            "token": sample_data["token"],
         }
 
     def __getitem__(self, idx: int) -> Dict[str, object]:

torchdrive/notebook.py

@@ -4,6 +4,7 @@
 import torch
 from IPython.display import display
 from torchvision.transforms.functional import to_pil_image
+from torchvision.utils import draw_bounding_boxes
 
 from torchworld.transforms.img import normalize_img, render_color
 
@@ -20,3 +21,28 @@ def display_color(x: torch.Tensor) -> None:
     Renders a [w, h] grid into colors and displays it.
     """
     display(to_pil_image(render_color(x)))
+
+
+def display_bboxes(img: torch.Tensor, bboxes: object, threshold: float = 0.5) -> None:
+    """
+    Displays bounding boxes in mmcv format on the provided image.
+    """
+    tboxes = []
+    labels = []
+    for i, box in enumerate(bboxes[0]):
+        if len(box) == 0:
+            continue
+        if not isinstance(box, torch.Tensor):
+            box = torch.from_numpy(box)
+        p = box[:, 4]
+        valid = p > threshold
+        box = box[valid]
+        tboxes.append(box[:, :4])
+        labels += [str(i)] * len(box)
+
+    tboxes = torch.cat(tboxes, dim=0)
+
+    img = normalize_img(img.float())
+    img = (img.clamp(min=0, max=1) * 255).byte()
+
+    display(to_pil_image(draw_bounding_boxes(image=img, boxes=tboxes, labels=labels)))

torchworld/structures/cameras.py

@@ -6,7 +6,7 @@
 
 import math
 import warnings
-from typing import Any, Dict, List, Optional, Self, Sequence, Tuple, Union
+from typing import Any, Dict, List, Optional, Sequence, Tuple, TYPE_CHECKING, Union
 
 import numpy as np
 import torch
@@ -16,6 +16,9 @@
 
 from torchworld.transforms.transform3d import Rotate, Transform3d, Translate
 
+if TYPE_CHECKING:
+    from typing import Self
+
 
 # Default values for rotation and translation matrices.
 _R: torch.Tensor = torch.eye(3)[None]  # (1, 3, 3)
@@ -366,7 +369,7 @@ def transform_points_screen(
 
     # pyre-fixme[14]: `clone` overrides method defined in `TensorProperties`
     #  inconsistently.
-    def clone(self) -> Self:
+    def clone(self) -> "Self":
         """
         Returns a copy of `self`.
         """

torchworld/structures/grid.py

@@ -1,13 +1,16 @@
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
-from typing import Optional, Self, Tuple, TypeVar, Union
+from typing import Optional, Tuple, TYPE_CHECKING, TypeVar, Union
 
 import torch
 from pytorch3d.structures.volumes import VolumeLocator
 
 from torchworld.structures.cameras import CamerasBase
 from torchworld.transforms.transform3d import Transform3d
 
+if TYPE_CHECKING:
+    from typing import Self
+
 T = TypeVar("T")
 
 
@@ -17,13 +20,13 @@ class BaseGrid(ABC):
     time: torch.Tensor
 
     @abstractmethod
-    def to(self, target: Union[torch.device, str]) -> Self:
+    def to(self, target: Union[torch.device, str]) -> "Self":
         ...
 
-    def cuda(self) -> Self:
+    def cuda(self) -> "Self":
         return self.to(torch.device("cuda"))
 
-    def cpu(self) -> Self:
+    def cpu(self) -> "Self":
         return self.to(torch.device("cpu"))
 
     @property